Glossary · Analytics

Golden query set

A curated set of queries with known-good results, used to score ranking changes offline.

The set represents how real users search: head terms, long-tail, typos, zero-result cases. Each query carries judgements — which documents are relevant, which are not — reviewed by a person.

Every ranking or synonym change is scored against the set before deployment. If mean nDCG drops past a threshold, the change is flagged for review rather than silently reaching production.

Related terms

nDCG A/B testing

Test AACSearch on your own problem

The best way to evaluate a search platform isn't reading one more comparison table.

Upload your own data.
Run real queries.
Look at the Query Trace.
Compare relevance.
Work out the cost.

Only then make the decision.

Start free Open Playground See the documentation